Modified SETI MB CUDA + opt AP package for full GPU utilization

Message boards : Number crunching : Modified SETI MB CUDA + opt AP package for full GPU utilization
Message board moderation

To post messages, you must log in.

Previous · 1 . . . 9 · 10 · 11 · 12 · 13 · 14 · 15 . . . 25 · Next

AuthorMessage
Profile Vipin Palazhi
Avatar

Send message
Joined: 29 Feb 08
Posts: 286
Credit: 167,386,578
RAC: 0
India
Message 849549 - Posted: 5 Jan 2009, 4:57:54 UTC - in response to Message 849407.  

Total GPU memory 939196416 free GPU memory 143332864

It seems your GPU has 1GB of memory but only ~100MB were free at start of task. So it said "out of memory".
Did you run any 3D graphic while using CUDA ? Why so low free memory ?

For another task:
Total GPU memory 939196416 free GPU memory 358807552

Again, where rest of GPU memory ?.... What driver do you use?

Please, try to answer of these questions and try to crunch few more CUDA tasks after OS reboot. It's pretty interesting case :)

I have the latest driver installed - 180.48. This rig is set up as a dedicated cruncher, so no other programs were running. As I mentioned, I restarted the system and still the same issue persisted.

Whats the error message you getting if you use the original version?
And second question: What is written in the first line of the script you using?

I dont exactly remember the error. Cant check the script as I am in the office at the moment, but the message was something like this - (34, 1) Could not find the file specified.
______________


ID: 849549 · Report as offensive
Profile Neil Blaikie
Volunteer tester
Avatar

Send message
Joined: 17 May 99
Posts: 143
Credit: 6,652,341
RAC: 0
Canada
Message 849554 - Posted: 5 Jan 2009, 5:15:17 UTC

I have just aborted a whole bunch of CUDA work as it was likely to error out. Am reverting back to last version of BOINC and not going to use CUDA at all until some of the bugs are ironed out.
ID: 849554 · Report as offensive
Profile Dirk Sadowski
Volunteer tester

Send message
Joined: 6 Apr 07
Posts: 7105
Credit: 147,663,825
RAC: 5
Germany
Message 849576 - Posted: 5 Jan 2009, 6:45:16 UTC


..worst case.. ..worst case..

I hope if two '-9 result_overflow'-error from CUDA will compare and for well diagnosed.. this WU will be saved somewhere and will send out to normal CPU-rigs later again!

..maybe there was the WOW-signal.. ?!?!

I thought we make/support this project for to find ET ?!?!

Not to get Credits for nonsense!


It make me sad.. maybe because of the 'nice' CUDA we have not noticed the WOW-signal ?


They Berkeley-crew is informed?

They have a solution for to eliminate this worst case?

ID: 849576 · Report as offensive
Profile Raistmer
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 16 Jun 01
Posts: 6325
Credit: 106,370,077
RAC: 121
Russia
Message 849599 - Posted: 5 Jan 2009, 9:02:23 UTC - in response to Message 849576.  


They Berkeley-crew is informed?

Yes.

They have a solution for to eliminate this worst case?

Don't know. But app debugging seems go, so sooner or later we will have more correct CUDA MB version....
ID: 849599 · Report as offensive
Profile Raistmer
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 16 Jun 01
Posts: 6325
Credit: 106,370,077
RAC: 121
Russia
Message 849600 - Posted: 5 Jan 2009, 9:06:26 UTC - in response to Message 849549.  

Total GPU memory 939196416 free GPU memory 143332864

It seems your GPU has 1GB of memory but only ~100MB were free at start of task. So it said "out of memory".
Did you run any 3D graphic while using CUDA ? Why so low free memory ?

For another task:
Total GPU memory 939196416 free GPU memory 358807552

Again, where rest of GPU memory ?.... What driver do you use?

Please, try to answer of these questions and try to crunch few more CUDA tasks after OS reboot. It's pretty interesting case :)

I have the latest driver installed - 180.48. This rig is set up as a dedicated cruncher, so no other programs were running. As I mentioned, I restarted the system and still the same issue persisted.


Free GPU mem amount on my host stays the same even after driver crash and your 2 results have different free GPU mem amounts both much smaller than total GPU mem available... Could you do few more resulst and post links on them here, please? Maube it's worth to try what will be with another driver version too....
ID: 849600 · Report as offensive
Profile Byron S Goodgame
Volunteer tester
Avatar

Send message
Joined: 16 Jan 06
Posts: 1145
Credit: 3,936,993
RAC: 0
United States
Message 850611 - Posted: 7 Jan 2009, 23:00:05 UTC
Last modified: 7 Jan 2009, 23:15:11 UTC

Had my first hiccup on an AP while running CUDA. Not sure it has anything to do with CUDA though, but since I'm running CUDA 24/7 I guess it potentilly could. Anyway, here's the task details.

Apparently this problem "In ap_fileio.cpp, Statefile::Read, statefile is 0'd, trying again:" kept coming up after it got about 48% of the way thru the task, and then ended with "Statefile::Read: 1 iterations with 0'd statefile.". At some point after that my pc rebooted, which I thought was due to a CUDA task (God knows it's happened before) but when Boinc was started again there was no error on a CUDA task before and after the reboot, but the AP was reset to 0% progress. I let the task run again and it finished without incident, also while still running CUDA MB.

AMD64 3800+
XP Home SP2
8500GT
180.60
ID: 850611 · Report as offensive
rascal
Avatar

Send message
Joined: 26 Apr 07
Posts: 18
Credit: 929,874
RAC: 0
Serbia
Message 850654 - Posted: 8 Jan 2009, 1:23:32 UTC

A question. Wher can i configurate my pc to use 4+1 cores? Can't be done in computing preferencess on my boinc acount page, the number of usable cores is already 16 and utilization of processors is 100%. So where do i change this? By the way does anyone have the answer to the ocasionaly crashing of video drivers?
"Mislim dakle postojim dakle kenjam non-stop"
ID: 850654 · Report as offensive
rascal
Avatar

Send message
Joined: 26 Apr 07
Posts: 18
Credit: 929,874
RAC: 0
Serbia
Message 850656 - Posted: 8 Jan 2009, 1:35:43 UTC

"I suspect the RAC could reach 8000 which
would be quite something for a q6600/8800gt based system. OC'd to 3.05 Ghz."

What's the word on RAC? My numbers don't seem to impresive. I've got Sparkle Calibre's 9600GT that's bought with OC-ing already done. But as i told you already the numbers aren't that impressive, in fact they are the same as 2 of my cores working on 2,5 GHz (not clocked Intel Q9300). Is this normal or am i missing something? In the meantime processor OC-ed to 3,43 GHz.

Help greatly needed and apreciated.
"Mislim dakle postojim dakle kenjam non-stop"
ID: 850656 · Report as offensive
rascal
Avatar

Send message
Joined: 26 Apr 07
Posts: 18
Credit: 929,874
RAC: 0
Serbia
Message 850687 - Posted: 8 Jan 2009, 2:57:27 UTC - in response to Message 850677.  

Question. Do i need to create that .xml file, because i do not know how to do that?

Help.
"Mislim dakle postojim dakle kenjam non-stop"
ID: 850687 · Report as offensive
Profile Voyager
Volunteer tester
Avatar

Send message
Joined: 2 Nov 99
Posts: 602
Credit: 3,264,813
RAC: 0
United States
Message 850691 - Posted: 8 Jan 2009, 3:06:45 UTC

Well i should have a board Fri.I've cleared the catche on my Pd and am ready to try this, then wheen I've got a handle on it switch to the quad. Is the stock cuda app more stable than the opp app? I thought I may start with the vanilla , better , no?
ID: 850691 · Report as offensive
Profile Byron S Goodgame
Volunteer tester
Avatar

Send message
Joined: 16 Jan 06
Posts: 1145
Credit: 3,936,993
RAC: 0
United States
Message 850697 - Posted: 8 Jan 2009, 3:14:57 UTC - in response to Message 850687.  

Question. Do i need to create that .xml file, because i do not know how to do that?

Help.

Open note pad, copy and paste

<cc_config>
<options>
<ncpus>5</ncpus>
</options>
</cc_config>

Save it as cc_config.xml

Make sure when you save it, to save it as "All files" under the file name

Put it in C:\ProgramData\BOINC

In Boinc Manager under the Advanced Menu click on read config file
ID: 850697 · Report as offensive
rascal
Avatar

Send message
Joined: 26 Apr 07
Posts: 18
Credit: 929,874
RAC: 0
Serbia
Message 850699 - Posted: 8 Jan 2009, 3:17:05 UTC

OK. Made the cc_config.xml file with the said instructions

<cc_config>
<options>
<ncpus>5</ncpus>
</options>
</cc_config>

placed it in "C:\ProgramData\BOINC" folder. Exited BOINC before copying the file to folder, than booted BOINC up again, but rosetta@home still using 3 cores and SETI@home still just the CUDA.
Am i doing something wrong.

By the way made the .xml file by copying the text to notepad and the just saving it as an .xml file (used the UTF-8 encoding)
"Mislim dakle postojim dakle kenjam non-stop"
ID: 850699 · Report as offensive
Profile Byron S Goodgame
Volunteer tester
Avatar

Send message
Joined: 16 Jan 06
Posts: 1145
Credit: 3,936,993
RAC: 0
United States
Message 850701 - Posted: 8 Jan 2009, 3:22:25 UTC - in response to Message 850699.  
Last modified: 8 Jan 2009, 3:25:25 UTC

Try the instruction "in Boinc Manager under the Advanced Menu click on read config file". Could be you didn't completely shut down Boinc
ID: 850701 · Report as offensive
rascal
Avatar

Send message
Joined: 26 Apr 07
Posts: 18
Credit: 929,874
RAC: 0
Serbia
Message 850712 - Posted: 8 Jan 2009, 3:44:06 UTC

after making the file restarted the whole computer, but nothing rosetta 3cores seti cuda core and that is it
"Mislim dakle postojim dakle kenjam non-stop"
ID: 850712 · Report as offensive
rascal
Avatar

Send message
Joined: 26 Apr 07
Posts: 18
Credit: 929,874
RAC: 0
Serbia
Message 850713 - Posted: 8 Jan 2009, 3:47:01 UTC

after suspending rosetta seti 3 cores + Cuda core
"Mislim dakle postojim dakle kenjam non-stop"
ID: 850713 · Report as offensive
rascal
Avatar

Send message
Joined: 26 Apr 07
Posts: 18
Credit: 929,874
RAC: 0
Serbia
Message 850716 - Posted: 8 Jan 2009, 3:49:56 UTC

So still not utilazing 100% of my cpu, he just ingaged astropulse aplication automatically, which i've disingaged in seti preferences. Anymore ideas?
"Mislim dakle postojim dakle kenjam non-stop"
ID: 850716 · Report as offensive
rascal
Avatar

Send message
Joined: 26 Apr 07
Posts: 18
Credit: 929,874
RAC: 0
Serbia
Message 850718 - Posted: 8 Jan 2009, 3:53:09 UTC

funny thing i forgot to mention. if i attach another project say GPUGrid it does work without the cc_config file, rosetta uses 4 cores and seti uses the CUDA, the minute you suspend GPUGrid it goes back to it's old ways. it has something to do with the share ratio being 33.33% to a project.
"Mislim dakle postojim dakle kenjam non-stop"
ID: 850718 · Report as offensive
Profile Byron S Goodgame
Volunteer tester
Avatar

Send message
Joined: 16 Jan 06
Posts: 1145
Credit: 3,936,993
RAC: 0
United States
Message 850723 - Posted: 8 Jan 2009, 4:01:37 UTC - in response to Message 850718.  
Last modified: 8 Jan 2009, 4:02:45 UTC

When you saved the cc config file the first time, did you save it as a text file or all files? Unless it's the file format used to save it I'm not seeing any reason why it shouldn't work.


funny thing i forgot to mention. if i attach another project say GPUGrid it does work without the cc_config file, rosetta uses 4 cores and seti uses the CUDA, the minute you suspend GPUGrid it goes back to it's old ways. it has something to do with the share ratio being 33.33% to a project.


I'm not sure I understand what you're saying. You're only attached to two projects, why wouldn't you have it set to 50% each in your settings.
ID: 850723 · Report as offensive
rascal
Avatar

Send message
Joined: 26 Apr 07
Posts: 18
Credit: 929,874
RAC: 0
Serbia
Message 850725 - Posted: 8 Jan 2009, 4:12:43 UTC

It is 50% to 50%, what i was talking about is when i added another project, a 3rd one.

Not really sure how i saved cc_config. I'll do it again as an "All Files" save.
"Mislim dakle postojim dakle kenjam non-stop"
ID: 850725 · Report as offensive
rascal
Avatar

Send message
Joined: 26 Apr 07
Posts: 18
Credit: 929,874
RAC: 0
Serbia
Message 850726 - Posted: 8 Jan 2009, 4:17:42 UTC

Saved as an "All Files" save option, with the given text in the file, copied in the "C:\ProgramData\BOINC" folder after shuting down BOINC and still just 3 cores and the CUDA.

God damn this is pissing me off!!!!!!!!!!

could it be the encoding of the text file. I used the UTF-8 option?
"Mislim dakle postojim dakle kenjam non-stop"
ID: 850726 · Report as offensive
Previous · 1 . . . 9 · 10 · 11 · 12 · 13 · 14 · 15 . . . 25 · Next

Message boards : Number crunching : Modified SETI MB CUDA + opt AP package for full GPU utilization


 
©2025 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.